At Data Science Dojo, I developed and owned services for Ejento AI, a production enterprise AI agent platform built with LangChain and LangGraph on Azure microservices. I worked on billing and quota controls, including metering integration and quota validation before LLM calls.
I owned the Guardrails service, improving response accuracy through LLM-generated guardrail responses and context-aware validation across multi-turn conversations. I also fixed database connection issues and vulnerabilities.
At ASF Tech Partners, I developed an AI-powered business onboarding platform for UAE company registration and engineered a RAG-based conversational AI with LangChain and Google Gemini. I also built a full-stack application with Supabase, authentication, role-based access control, and admin dashboards.
My projects include forecasting commodity prices with LSTM models, building a backend for the Karavaan social media application, and developing explainable image forensics and Urdu text summarization systems. At Pakistan Agriculture Research (PAR), I cleaned commodity datasets and developed forecasting models that achieved 92% predictive accuracy on historical data.
